Documentation

Connectivity Data

Synaptic connectivity on Virtual Fly Brain — what an edge is, which connectomes contribute one, and how to query across them.

15 Aug 2026connectomeEMconnectivity

Connectivity is derived data: it comes out of the EM datasets, but VFB holds and queries it separately from the images. Loading it into one knowledge graph, with both ends of every edge classified against the Drosophila Anatomy Ontology, is what makes it possible to ask a question once and have it answered across every connectome at the same time, rather than once per resource in each resource’s own vocabulary.

488,111
neurons with connectivity
34.7 M
neuron-to-neuron connections

What an edge means

A VFB connectivity edge records that a reconstruction reported synapses between two neurons, and how many. It is evidence from one dataset, not a biological fact: two neurons may be connected in the animal and unconnected in a partial reconstruction, and a weak edge may be a segmentation artefact. Treat every count as attributable to the connectome it came from.

VFB holds four kinds of connection:

FromToWhat it is
Neuron (individual)Neuron (individual)The connectome proper — one reconstructed neuron synapsing onto another
Neuron (individual)Region (individual)Synapses a neuron makes within a named neuropil, used for region-level input and output profiles
Neuron (class)Muscle (class)Curated motor innervation, from the literature rather than from EM
Neuron (class)Sense organ (class)Curated sensory innervation, likewise

The first two are per-individual and come from EM. The last two are per-class and come from curation, so they carry no synapse counts and cover far fewer entities. The content report gives the current size of each.

Which connectomes contribute

DatasetVFB symbolCoverage
BANC v626BANCFull CNS, adult female
FlyWire v783 (FAFB)fwFull brain, adult female
male-CNS v0.9mcFull CNS, adult male
MANC v1.2.1mvFull VNC, adult male
Optic-lobe v1.0.1olOptic lobe, adult male
hemibrain v1.2.1hbPartial brain, adult female
FAFB (CATMAID)fafbFull brain, adult female — sparse reconstruction
L1 CNS (CATMAID)l1emFull CNS, first-instar larva

Versions matter: a connectome is re-released as proofreading progresses, neuron identifiers do not always survive between versions, and VFB states the version it holds. See EM dataset versioning.

Overlapping reconstructions

Several of these reconstruct the same tissue. The hemibrain, the sparse FAFB CATMAID reconstruction and FlyWire all cover adult female brain, and a connection present in more than one of them is one biological connection reported several times, not several connections. Summing partner counts across all datasets therefore inflates them.

The convention VFB and VFB_connect follow is to exclude hb and fafb from cross-dataset connectivity queries by default, leaving FlyWire as the adult female brain source. Change that when the question calls for it — comparing reconstructions against each other, for instance — but change it deliberately.

Querying it

Connectivity queries are available from the Term Info pane of any neuron or region, and programmatically through VFB_connect and the MCP server.

  • Partners of one neuron type. Ask for its inputs or outputs. The default weight threshold is 5 synapses; lower it only if you are prepared to look at the weak edges individually.
  • Between two neuron types. A pairwise query returns the edges between them in each contributing dataset separately, which is usually what you want to see.
  • For a region. Regions take a different set of queries from neurons. Ask a neuropil for the neurons presynaptic or postsynaptic in it, not for its connectivity partners.

Direction is easy to invert, so it is worth stating plainly: neurons presynaptic in a region have their presynaptic terminals there and are providing input to it; neurons postsynaptic in a region receive there. A query for the inputs to a region returns the presynaptic set.

What the numbers do and do not say

The neuron counts above are the neurons VFB holds connectivity for, which is not the number of neurons in a fly. Absence is likewise not evidence: a cell type with no connectivity on VFB may simply not be reconstructed in any dataset VFB has loaded, or may not yet be matched between the connectome’s own naming and the ontology. Where a connectome’s identifiers have not been mapped onto ontology classes, the neurons are still present and queryable individually.